G-Power: Architecture-level GPU Power Modeling with Aggregated Knowledge Foundations from Known GPUs

2026-08-07

AI Summary

This work addresses the limited accuracy of existing architectural-level GPU power models, which stems from their reliance on scarce microbenchmark training data and hinders effective optimization. To overcome this, the authors propose G-Power, a novel framework that, for the first time, integrates aggregated knowledge from multiple source GPUs with an attention mechanism. G-Power employs a three-stage pipeline—pre-training, attention-based knowledge aggregation, and fine-tuning on the target GPU—to enable effective cross-architecture knowledge transfer. This approach transcends the constraints of traditional microbenchmark datasets, substantially enhancing model generalization and prediction accuracy. Evaluated across four modern NVIDIA GPUs, G-Power achieves an average MAPE of 14% and a correlation coefficient (R) of 0.88, outperforming AccelWattch by reducing error by 22% and improving correlation by 0.36.

Abstract

Graphics Processing Units (GPUs) have been serving as critical computation resources for large-scale parallel computations. With increasing chip complexity, power efficiency has become an important design objective for modern GPUs. GPU power optimization relies on fast power evaluation, requiring architecture-level GPU power model. However, because of the time-consuming power label collection, only simple microbenchmarks are adopted for training. The limitation of microbenchmarks as training data incurs low accuracy for existing architecture-level GPU power models like AccelWattch. To address the limitation of microbenchmarks as training data, we propose G-Power, an architecture-level GPU power modeling framework that utilizes additional known GPU chips to provide additional knowledge. G-Power utilizes the aggregated knowledge foundation from additional known GPU chips and then performs fine-tuning on our target GPU. To provide foundations with additional known GPU chips and capture the similarity to utilize these foundations for fine-tuning, G-Power adopts a three-phase algorithm consisting of 1) pre-training with additional known chips, 2) attention-inspired aggregation, and 3) fine-tuning on our target GPU. We evaluate G-Power on four modern NVIDIA GPUs, demonstrating high accuracy. G-Power can achieve a low MAPE of 14% and a high correlation coefficient R of 0.88 on average, which are 22% lower MAPE and 0.36 higher R than AccelWattch.

Problem

• GPU power modeling
• architecture-level
• microbenchmarks
• power efficiency
• training data limitation

Innovation

• GPU power modeling
• knowledge aggregation
• transfer learning
• architecture-level modeling
• fine-tuning